index: support webarchive (.war) and mimehtml (.mhtml) formats

src/filters/rclwar

@@ -0,0 +1,54 @@
+#!/usr/bin/env python
+
+# WAR web archive filter for recoll. War file are gzipped tar files
+
+import rclexecm
+import tarfile
+
+class WarExtractor:
+    def __init__(self, em):
+        self.em = em
+
+    def extractone(self, tarinfo):
+        docdata = ""
+        try:
+            member = self.tar.extractfile(tarinfo)
+            docdata = member.read()
+            ok = True
+        except Exception, err:
+            self.em.rclog("extractone: failed: [%s]" % err)
+            ok = False
+        return (ok, docdata, tarinfo.name, rclexecm.RclExecM.noteof)
+
+    ###### File type handler api, used by rclexecm ---------->
+    def openfile(self, params):
+        self.currentindex = 0
+        try:
+            self.tar = tarfile.open(params["filename:"])
+            return True
+        except Exception, err:
+            self.em.rclog(str(err))
+            return False
+
+    def getipath(self, params):
+        ipath = params["ipath:"]
+        try:
+            tarinfo = self.tar.getmember(ipath)
+        except Exception, err:
+            self.em.rclog(str(err))
+            return (False, "", ipath, rclexecm.RclExecM.noteof)
+        return self.extractone(tarinfo)
+
+    def getnext(self, params):
+        tarinfo = self.tar.next()
+        if tarinfo is None:
+            #self.em.rclog("getnext: EOF hit")
+            return (False, "", "", rclexecm.RclExecM.eofnow)
+        else:
+            ret = self.extractone(tarinfo)
+            return ret
+
+# Main program: create protocol handler and extractor and run them
+proto = rclexecm.RclExecM()
+extract = WarExtractor(proto)
+rclexecm.main(proto, extract)

src/sampleconf/mimeconf

@@ -71,12 +71,13 @@ application/x-flac = execm rclaudio
 application/x-gnuinfo = execm rclinfo
 application/x-kword = exec rclkwd
 application/x-lyx = exec rcllyx
+application/x-mimehtml = internal message/rfc822
 application/x-perl = internal text/plain
 application/x-scribus = exec rclscribus
 application/x-shellscript = internal text/plain
 application/x-tex = exec rcltex
-text/x-tex = exec rcltex
 application/x-chm = execm rclchm
+application/x-webarchive = execm rclwar
 application/zip = execm rclzip;charset=default
 audio/mpeg = execm rclaudio
 audio/x-karaoke = execm rclkar
@@ -104,6 +105,7 @@ text/x-purple-log = exec rclpurple
 text/x-purple-html-log = internal text/html
 text/x-python = exec rclpython
 text/x-shellscript = internal text/plain
+text/x-tex = exec rcltex
 
 ## #############################################
 # Icons to be used in the result list if required by gui config

src/sampleconf/mimemap

@@ -112,6 +112,10 @@
 
 .fb2 = text/x-fictionbook
 
+.war = application/x-webarchive
+.mht = application/x-mimehtml
+.mhtml = application/x-mimehtml
+
 # A list of suffixes (name endings) that we don't want to touch at all.
 # Having these explicitely listed speeds things up a bit by avoiding
 # unneeded decompression or 'file' calls. File names still get indexed if